Project Summary In sub-Saharan Africa, the primary data for estimating HIV epidemic trends is HIV prevalence among pregnant women attending antenatal care (ANC). ANC sentinel surveillance has been conducted every one to two years since the early 1990s across the continent. A ?convenience sample? consisting of ten to twenty facilities in each country were selected to be sentinel sites in which a sample of pregnant women were tested for HIV each year. National estimates of HIV incidence and prevalence are created by fitting a mathematical model called ?EPP? to the HIV prevalence trend in these clinics and prevalence estimates from national household surveys. This makes strong assumptions that epidemic trends in a few non-randomly selected sentinel sites are representative of the national epidemic. This project will evaluate and improve these assumptions about how to extrapolate from sentinel sites to national trends, specifically aiming to (1) characterize the spatial representativeness of selected ANC sentinel sites across sub-Saharan Africa, (2) test for differences in HIV epidemic trends across sentinel sites, (3) develop and validate an approach to propagate uncertainty in epidemic estimates resulting from relying on a small number of sentinel site locations, and (4) demonstrate the impact of the these findings for estimates of HIV prevalence and incidence trends in sub-Saharan Africa. The primary data will be geo-located existing ANC sentinel surveillance data from countries across sub-Saharan Africa, and will also rely on remotely sensed population spatial covariates for population density and accessibility. Spatial and longitudinal statistical models will be used to test several hypotheses: H1.1: sentinel sites are disproportionately selected in areas of higher population density and more accessible to major roadways and urban centers; H1.2: population HIV prevalence is higher in areas surrounding selected sentinel sites; H2.1: HIV prevalence declined more in sites with higher prevalence; and H3.1: statistical uncertainty about historical incidence and prevalence is much greater than represented by current estimates that do not account for sentinel site selection. Taken together, the implications of these hypotheses may be that current interpretation of ANC surveillance data has resulted in systematically over-estimating peaks and declines in national HIV epidemics in sub-Saharan Africa and spuriously precise epidemic estimates that do not give adequate weight to more representative data sources, such as national household sero-surveys. Results of the project will be reported to the UNAIDS Reference Group on Estimates, Modelling, and Projections (http://www.epidem.org) and will result in improved methods for the UNAIDS EPP/Spectrum software used to generate official national estimates of HIV prevalence, incidence, and mortality in sub-Saharan Africa.